[
https://issues.apache.org/jira/browse/HBASE-4438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13108347#comment-13108347
]
Jesse Yates commented on HBASE-4438:
------------------------------------
I agree that those kind of tests still count as functional. When you are
injecting faults into it, I would also worry about whats going to happen on the
backend, when you are reusing the minicluster for multiple tests.
It would be cool if we can then break out short and long running functional
tests and run them as desired.
However, as Doug commented on the thread though, the mini cluster is used in a
lot of cases to do true functional testing (spin up mini cluster w/ a single
node, do some stuff, then shut it down). I feel like there should be a fair
number of non-daemon-interacting 'unit' tests which really shouldn't be
considered unit tests.
I don't have any experience tying in annotations to the build system, but I
would be interested to see how that would all work.
I feel like it still makes sense to break out integration tests, we just need
to be careful about how we do it. What do you think?
> Speed the build test phase
> --------------------------
>
> Key: HBASE-4438
> URL: https://issues.apache.org/jira/browse/HBASE-4438
> Project: HBase
> Issue Type: Umbrella
> Components: test
> Reporter: stack
>
> We're up to 2hrs 15mins running tests on jenkins which is kinda crazy. Akash
> started a thread over in dev on fixing this:
> http://search-hadoop.com/m/cZjDH1ykGIA/Running+UnitTests+before+submitting+a+patch&subj=Running+UnitTests+before+submitting+a+patch
> Jessie brings up http://maven.apache.org/plugins/maven-failsafe-plugin/ which
> looks like something we need to start using; having a tiering of our testing
> would make sense to me.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
